Objective To evaluate algorithms to predict insignificant prostate cancer at radical prostatectomy (RP). Methods Five hundred and fifty men (410 Caucasian, 100 African American [AA], and 40 others) with prostate-specific antigen (PSA) level <10 ng/dL, T1c, 12-core biopsy, and biopsy Gleason score 3 + 3 = 6 were categorized into training and validation sets. Six biopsy algorithms were tested for predicting insignificant (0.5 cm3, organ confined, and Gleason score ≤6) cancer at RP. Cancers incorrectly predicted to be insignificant were ranked into 4 groups of increasing aggressiveness. Results Original (Gleason score 6, PSA density ≤0.15, 2 positive cores, and maximum core involvement ≤50%) and modified Epstein criteria (Gleason score 6, PSA density ≤0.15, 2 positive cores, and unilateral cancer) had the highest negative predictive values - correct classification of insignificant cancer. Among cancers predicted to be insignificant in Caucasians, 29.9% cases using the original and 27% cases using the modified Epstein criteria had significant cancer at RP. However, more adverse findings at RP were misclassified as insignificant in only 3.5% and 2.2% of cases using the original and modified Epstein criteria, respectively. Of cancers predicted insignificant in AA men, 54.1% cases using the original and 51.6% cases using the modified Epstein criteria were misclassified as insignificant. Dominant anterior tumors were seen in 117 Caucasian (28.5%) and 44 AA men (44%). Conclusion The Epstein criteria maintain their accuracy in the modern era with extended biopsy sampling. The negative predictive values are lower in AA men, in part due to higher frequency of anterior tumors, where multiparametric magnetic resonance imaging should be recommended in AAs considering surveillance.
ASJC Scopus subject areas